Instructor:
Kelley Kidwell, PhD (University of Michigan)

Workshop Dates and Times:
Thursday, June 13th, 10:00am – 3:30pm ET
Friday, June 14th, 10:00am – 3:30pm ET

Workshop Format:
Two-Day Synchronous Online Workshop

If only we could take one pill and be cured of what ails us. Unfortunately, most diseases and disorders require ongoing, tailored treatment that adapts to an individual’s needs and responses. Yet, evidence for effective treatment is often produced at one point in time. That is, the gold standard of evidence, a randomized, controlled trial (RCT), most often investigates if an intervention is effective in a sample of individuals at one point in their disease or disorder progression. Although the classic RCT is valuable, we may be missing treatment synergies or antagonisms by not investigating more of the intervention process in the same individuals. These effective treatment guidelines that specify whether, how, when or for whom to alter treatment at critical decision points across the scope of care are called adaptive interventions or dynamic treatment regimens. Clinicians may use judgement and evidence from point-in-time RCTs to piece together these guidelines, but they may suffer from bias due to sample selection or confounding. One way to rigorously develop adaptive interventions is with sequential, multiple assignment, randomized trials (SMARTs).

A SMART is a type of multi-stage randomized trial design such that participants are randomized at least twice in sequence. Subsequent randomization and/or treatment assignment is based on response to previous treatment. Within a SMART design, several adaptive interventions. A SMART is motivated by the development of these adaptive interventions. Analytic methods estimate the effects of the adaptive interventions at the end of the trial.

SMARTs have been applied in a variety of health research areas including cancer, toxicology, mental health, HIV, weight loss, chronic pain, and tobacco cessation, education and in implementation science. The NIH, PCORI, and other funding agencies have funded many SMART designs in these areas. Most SMARTs employ individual randomization, however, SMARTs may also use cluster randomization. Cluster randomization is applied especially when considering adaptive interventions in implementation science or disease prevention. Methods exist for a wide variety of SMART designs considering many different outcome types.

This workshop will provide an introduction into adaptive interventions and describe how SMART designs can be used to develop high-quality adaptive interventions. The workshop will first introduce adaptive interventions providing definitions, breaking down their components, explain why and where they are relevant, and present examples across a variety of settings. Next, the workshop introduces SMART design. The design is defined and best illustrated through various case studies. Each of the (real) case studies includes specific characteristics that differentiate one from another and are used throughout the course. SMART designs are compared and contrasted to other similar. SMART design principles are discussed along with the typical primary, secondary and exploratory aims, especially considering the needs of grant proposal or trial protocol. The rest of the workshop focuses on a high level overview of analytic methods of a SMART design including analysis for primary, secondary and exploratory aims, along with power and sample size calculations. Each method is illustrated through the case studies and coding in R is provided.